F3-16000CL9T-6GBTDD & Gigabyte EX58-UD4P

[Eber]

System Specs:
i7 920(CO stepping)
Gigabyte EX58-UD4P(BIOS: F6 when I first got the memory, updated to F13 with same results)
previous memory: hyper-x 3gb 1866mhz 9-9-9-27 1.65v
new memory: G Skill Trident 6gb 2000mhz 9-9-9-24 2n 1.60v(F3-16000CL9T-6GBTDD)

Previously stable overclock(stable for 16+ hours in Prime 95):
190x21(turbo on)
Hyperthreading: off
C1E, C3/C6/C7, EIST: off
QPI: x36
Uncore: x17
RAM: x8(1520 mhz)
Timings: 8-8-8-24 1T(advanced timings on auto)
Performance mode: Standard
Load Line Calibration: on
Vcore: 1.30v
VTT: 1.335v
DRAM: 1.64v
All other voltages and settings are on auto.


I'm having trouble getting my new memory kit stable at anything over 1066 and when I tried the XMP profile my PC got stuck in an endless reboot loop(I had to use the clear CMOS button to get it to boot again). With everything at stock speed and auto voltages I was able to run prime 95 fine for a few hours until I stopped it.

I was hoping to run the new kit at 1900mhz but I thought I would try my old settings(other than dram at 1.60v) as a starting point(I know, different memory but the new kit is rated for higher speeds and tighter timings at lower volts), at those settings my PC froze in about 5 minutes running prime. I loosened the timings to 9-9-9-24 2N at which point it would freeze within 10-15 minutes, I tried bumping up the VTT to 1.355v and the dram to 1.62v but neither helped at all. At this point I decided to update my BIOS since some of the newer ones listed memory compatibility enhancements, then I tried different VTT and DRAM voltages again. Next I tried dropping the ram multi to 6x(uncore to 13x) to run it at 1140mhz, after that prime would freeze after about 45 minutes though that was after upgrading the BIOS so I might need to re-adjust vcore or something.

With my old kit I could run it at 1866mhz using the XMP profile but I think the highest I got it running with manual settings was about 1650mhz, I think it was an issue with advanced timing or IOH/PLL voltages as I tried liberal amounts of DRAM voltage(1.70v) and VTT voltage(1.6)

I thought I should mention that I'm not using the ram cooler that came with the kit and I'm stress testing under XP(32 bit) because it's an old install that I don't mind corrupting, I plan on testing and tweaking for win 7 after I get the OC stable in XP. I don't think that those things should be an issue but I thought I should mention them in case I'm wrong.

I haven't run memtest+ or tested individual sticks yet as I'm hoping I just need a voltage or advanced timing tweak. I bought the memory from newegg about 2 weeks ago so I'm hoping to straighten this out in the next week or so in case I need to return them.

I apologize if I rambled too much and thanks in advance for any help.

[Tradesman]

May want to run memtest on each stick individually

[Eber]

Since posting that I actually ran memtest86+ but with all of the ram. It passed 2 passes at 1140 and 2 passes at 1520, both at at 9-9-9-24 2 1.60v. Should I try them individually with the xmp profile since the xmp profile wouldn't even boot with all 3?

[Tradesman]

Might give it a try

[Eber]

So I took 2 sticks out leaving the one in the first slot in, booted into the bios and set all of the clocks and voltages to auto and then enabled the xmp profile. I rebooted and it booted into windows normally, I ran prime for a few minutes without issue.

Then I decided to see if the other two sticks would boot with xmp profile with only one in. I turned off the computer and power supply, put one of the other sticks of ram in the first slot, I then tried to boot the computer and it was endlessly rebooting(fans would spin up for a second and then everything would turn off for a second, then repeat...). I tried the last stick and it wouldn't post either(rebooting).

I tried the stick that had worked and it wouldn't post so I reset the cmos, it then booted into windows normally. I rebooted into the bios and enabled the xmp profile and it wouldn't post again. I tried the other 2 sticks(clearing the cmos, and then enabling xmp with each stick) in case I had mixed them up but they still wouldn't post.

I noticed that the lower cpu multiplier from the xmp profile was sticking after clearing the cmos and I wondered if some of the power saving features helped that I still had disabled when the one stick worked with the xmp profile. I tried each stick by clearing the cmos, loading optimized defaults in the bios, disabling hyperthreading c1e and eist, setting performance mode to standard, and then enabling the xmp profile but they still wouldn't post after the xmp profile was enabled.

I tried one last time with the stick that worked once booting into windows between each set of changes in the bios(optimized defaults, disabling ht c1e eist & set perf mode, enable xmp), it still wouldn't post.

I put my old 3gb kit in, disabled power saving features and such, enabled the 1866mhz xmp profile and it booted and ran prime for a few minutes fine. I then tried to set up the overclock mentioned in the first post to see if it was stable with the new bios. Prime froze my pc after 9 minutes which is farther than I got with the new kit at 1540mhz 8-8-8-24 1t, the only difference being dram at 1.60v(new) vs 1.64(old). I bumped vcore up to 1.3125v and prime ran for over an hour without any issues. That's not 24/7 stable but it is close.

I might see if I can get the new kit stable with my overclock now that I know I need more vcore with the new bios but I would like to know what voltage range I should be playing with for dram and vtt. I would also like to know if I should try different advanced timings or voltages for ioh, pll, etc... I would like to get the memory running at 1900mhz but I would be satisfied to get 1540mhz with tight timings and low volts. I'm running short on time to return the memory so I would appreciate any help with settings to try and get it working before I have to return it.

Again I apologize if I rambled too much I just didn't want to miss any possibly important details.

[GSKILL TECH]

Are you adjusting the QPI/VTT Voltage any? This is what your concentration should be on. Also, search x58a on this forum, there are many previous users with this mobo and memory combination.

Thank you
GSKILL TECH

[Eber]

As I mentioned in my first post I did try bumping up the vtt some(1.335 to 1.355 at 1520mhz 9-9-9-24 2) which isn't much but it had zero affect on stability. Before I posted I spent a couple hours searching the forum for posts mentioning my motherboard and memory but none of the threads seemed to have any information related to my problem.

In the past I have been very happy with G Skill memory and have heard good things about customer service but this experience has been the exact opposite on both counts. This kit has been a nightmare to work with and after spending a good deal of time to post detailed descriptions of the situation I have received terse generic responses that leave most of my questions unanswered. I don't mean to be rude I'm just getting really frustrated.

[Eber]

So after trying the kit again with the extra vcore I need with new bios I had a little more success and I'm a little less frustrated than when I posted last. I got 1520mhz 8-8-8-21 1t somewhat stable with vtt at 1.335v and dram at 1.60v, I also got 1900mhz 9-9-9-24 2 to boot with vtt at 1.595v and dram at 1.60v but it was crashing within a couple minutes of starting prime until I tried setting the uncore multi at 20x instead of 21x(21x uncore has always worked better with 10x mem?).

Even with the uncore at 20x it still crashed after 49 minutes in prime at 1900mhz 9-9-9-24 2 so it's not stable but it seems like like it should be possible to get it stable. I tried loosening the tRAS to 27, I'm going to test that next.

I was going to list the automatic settings listed for advanced timings to see if they looked right but I noticed that the bios was listing the auto setting for tRFC as 60 and CPU-Z shows it at 106. Should I set any of the advanced timings manually?

I noticed that the xmp profile raises "dram termination" and "ch-a data vref" but it doesn't say what it raises them to, should I be changing those?

Last but not least is 1.595v vtt relatively safe? I'm pretty sure my last bios listed anything above 1.5 as dangerous but the new one lists anything above 1.6 as dangerous and intel specs say 1.35v is the max officially but unofficially intel just says to keep vtt within .5v of dram, it's all a little confusing.

[GSKILL TECH]

This is high performance memory, so overclocking is required. If you are unsure of how to do this and what is involved, research should be done to be more familiar.

You have an i7-920 C0, so you should know that most are not capable of DDR3-2000. If at all possible, VTT would be 1.60V+. What I am saying in the previous post is that you should not worry about all the other settings, the VTT is what's going to determine stability. You should see what the highest your CPU can do first, then tweak other settings. The main difficulty is the VTT Voltage.

For uncore, it is just double memory, so no real need to adjust that. Intel specs state double is more than enough.

tRFC can be left AUTO, 106 is more correct than 60.

You should not need to adjust DRAM Term and vREF, those will be handled by the motherboard.

QPI/VTT 1.595V is perfectly fine, as long as the CPU cooler is doing its job. They state 1.35V max because they don't want people to overclock and fry the CPU. Advanced/extreme overclockers know the limitation, which is why we allow people to set it that high. Any other company will not help you with this, it is not their duty as overclocking is not supported. We make overclocking modules, we overclock ourselves, customer support is our priority, so it is only natural that we share our knowledge with those that would like to know.

If you are willing to work with us, we can help and guide you, but please be patient because it is not our duty that you understand how to overclock. We love to share knowledge and that's why we do end up teaching people what we can, when we can.

Thank you
GSKILL TECH

[Eber]

Thank you for a more detailed response with some useful information. I guess I should have mentioned that I am familiar with overclocking, I have been running that 3.99ghz oc I mentioned for two years and have been overclocking for about 10 years now. I'm not an expert overclocker and I did most of my research regarding overclocking on the x58 platform when it was still fairly new and the affects and consequences of some the voltages(vtt, ioh, pll) were still relatively unknown. I asked about some of those settings because I thought you might have more up to date info but I can see how it might seem like I don't know much about overclocking. Now I AM a newb when it comes to advanced timings, that stuff seems like half voodoo to me.

I only mentioned the uncore thing because it's the exact opposite of my previous experience with this board, in the past 2xmem+1 was stable and 2xmem was not. I thought it might be a clue to my problems or at least be useful in the future for someone with my board in the same situation.

I did a little research before seeing your response and it seems the general consensus on vtt is that 1.6v or less is fine and that a little more might be ok but that's still up for debate. I don't really feel comfortable setting vtt above 1.6v, is there anything else I can do to get that last little bit of stability at 1900mhz or is more vtt the only thing?

I had another problem with the memory today but it helped me figure out part of my problems with xmp. I left my memory at 1900mhz and when I started my PC today it wouldn't post(reboot loop) like it had been doing with the xmp profile enabled, I had to clear the cmos to get it to boot. After a few minutes I booted into the bios and set the oc back up with the memory at 1900mhz and it booted up just fine. This got me thinking so I tried the xmp profile while the computer was warmed up and it worked, cpu-z confirmed that it was running at the proper speed and timings. I confirmed that it was indeed a cold boot issue by turning off my PC for 15 minutes and then tried to turn it on and it wouldn't post. My old 1866mhz kit does not have this cold boot issue using the xmp profile.

Since my old memory sets vtt to 1.5v I'm guessing that maybe my motherboard doesn't like the higher vtt or that maybe the new kit is defective, any thoughts?

I should add that I did an overnight run of prime with my old memory kit to test my oc with the new bios and it is 100% stable so I really am just trying to get the memory dialed in at this point.

[Tradesman]

Could you post your current BIOS settings, so we can see if anything looks off

[Eber]

Wow, I wasn't expecting a reply over the weekend. The settings I used to try and get the mem at 1900mhz are listed below, any settings in MIT not listed are on default. With these settings the computer failed after 49 minutes of prime.

CPU Clock Ratio=21x(with turbo enabled on my mobo it treats it like a regular multiplier)
Intel Turbo Boost=Enabled
CPU Multithreading=Disabled
C1E=Disabled
C3/C6/C7=Disabled
EIST=Disabled
QPI Clock Ratio=x36
Uncore Clock Ratio=x20
Bclk Control=Enabled
Bclk Frequency=190
Performance Enhance=Standard
System Memory Multiplier=10.0
DRAM Timing Selectable=Quick
CAS=9
tRCD=9
tRP=9
tRAS=24
Command Rate=2
Load Line Calibration=Enabled
CPU Vcore=1.31250v
QPI/VTT Voltage=1.595v
DRAM Voltage=1.600v

Thanks for taking a look.

[Tradesman]

Might move vCore up to 1.325 and QPI/VTT to 1.6, try that and I'll check a couple things and be back this in a few hours

[Eber]

I'll test it with vcore at 1.325. I can't set vtt to 1.6v with this board, it's either 1.595v or 1.615 and that's getting into the range I'm not comfortable with.

Any ideas about the cold boot issue, if I can't get around that then it doesn't matter if I can get it stable.

Thanks again

[Tradesman]

Do you have the C1E, SpeedStep, etc power savings in the BIOS disabled